Date:         Wed, 25 Jun 1997 17:38:59 GMT
Reply-To:     David Nichols <nichols@SPSS.COM>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From:         David Nichols <nichols@SPSS.COM>
Organization: SPSS, Inc.
Subject:      Re: Homogeneity of Variance and Repeated measures

In article <5origb$hjq$1@news.fsu.edu>,
James D. Young <jyoung@mailer.fsu.edu> wrote:
>This is more of a stats question than an SPSS question, but here goes...
>
>When running a repated measures ANOVA, SPSS spits out both univariate
>homogeneity of variance tests for each level of my repeated factor (e.g.,
>Bartlett-Box) and Multivariate test for Homogeneity of Dispersion matrices
>(e.g., Boxs M). When reporting my Within subjects effects, which
>homogeneity factor am I concerned with? Frankly, I have no idea what the
>latter test even is. Should I report both?
>
>Your wisdom would be greatly appreciated
>
>James
 
The Box test is really more appropriate, because the assumption of
equality of covariance matrices is what's necessary for the tests
done with repeated measures to be strictly valid (this is necessary,
not sufficient, for the averaged univariate approach tests, which
also require the circularity or sphericity condition to be met).
Unfortunately, Box's M test isn't particularly robust, but I don't
know of an accepted alternative.
 
--
-----------------------------------------------------------------------------
David Nichols             Senior Support Statistician              SPSS, Inc.
Phone: (312) 329-3684     Internet:  nichols@spss.com     Fax: (312) 329-3668
-----------------------------------------------------------------------------